Download The Mere-Reaction Effect: Even Nonpositive and

The Mere-Reaction Effect: Even
Nonpositive and Noninformative
Reactions Can Reinforce Actions
5
CHRISTOPHER K. HSEE
YANG YANG
BOWEN RUAN
Prior research indicates that a stimulus can reinforce an action if the stimulus is a
reward (i.e., a priori positive) or carries useful information. The current research finds
that if a stimulus is perceived as a reaction to an action, it can reinforce the action
even if the stimulus is a priori nonpositive and noninformative. Mere reactions are
reinforcing. Specifically, eight experiments, including a field experiment, demonstrate
that individuals are more likely to repeat an action (e.g., inserting money in a donation
box or typing a message in a textbox) if the action is followed by a stimulus (e.g., the
emission of a sound or the flash of an image) than if it is not, even if the stimulus is a
priori negative (e.g., an annoying sound or an aversive image) and carries no useful
information. Moreover, the effect just described will occur only if the stimulus is contingent on (immediately follows) the action and perceived as a reaction to the action.
Finally, by serving as a reaction, an a priori nonpositive stimulus can become positive.
The present work yields theoretical implications for stimulus–response relationships
and practical implications for designs of consumer products and loyalty programs.
10
15
20
Keywords: reaction, motivation, reinforcement, valence, contingency, control,
heuristic, overgeneralization, product design
25
A
tip jar sits on the bar of a restaurant, and invites customers to insert money into it. Consider four alternative editions of the tip jar: The basic edition would neither
emit any sounds nor provide any useful information when
one inserts money into it. The positive-valence edition
would emit a delightful tone every time one inserts money
into it. The useful-information edition would display some
useful information every time one inserts money, such as
how big one’s tip is relative to others, how many servers
will share the tip, and so on. Finally, the mere-reaction edition would merely emit a sound every time one inserts
money into it, and the sound is negative in valence, such as
an annoying beep.
Suppose a customer tips (puts money in the jar) at least
once. Then, relative to the basic edition, in which edition(s)
will the customer be more likely to repeat inserting money,
and in which edition(s) will she be less likely?
The existing literature yields two obvious predictions.
The first is that one will be more likely to repeat inserting
money in the positive-valence edition than in the basic edition. This is due to operant conditioning, according to
which the presence of an a priori positive stimulus (i.e., a
reward) following a behavior reinforces the behavior, and
Christopher K. Hsee (e-mail: [email protected]) is the
Theodore O. Yntema Professor of Behavioral Science and Marketing,
Booth School of Business, University of Chicago, Chicago, IL 60637.
Yang Yang (e-mail: [email protected]) is an assistant professor of marketing, Warrington College of Business Administration,
University of Florida, Gainesville, FL 32611. Bowen Ruan is a doctoral
student in marketing, Wisconsin School of Business, University of
Wisconsin-Madison, Madison WI 53706. The authors thank the
Templeton Foundation and their respective schools for research support,
and thank the review team and the following individuals (in alphabetical
order) for helpful suggestions: Cindy Cai, Jeff Galak, Carey Morewedge,
Luxi Shen, and Oleg Urminsky. Correspondence should be addressed to
either Christopher K. Hsee or Yang Yang.
Gita Johar served as editor and Lisa Bolton served as associate editor for
this article.
Advance Access publication 11 May 2015
C The Author 2015. Published by Oxford University Press on behalf of Journal of Consumer Research, Inc.
V
All rights reserved. For permissions, please e-mail: [email protected] Vol. 0 2015
DOI: 10.1093/jcr/ucv022
1
30
35
40
45
2
5
10
15
20
25
30
35
40
JOURNAL OF CONSUMER RESEARCH
the presence of an a priori negative stimulus (i.e., a punishment) following a behavior weakens the behavior (Skinner
1953, 1957; Thorndike 1898). A delightful tone is an a priori positive stimulus and hence will reinforce the tipping
behavior.
The other obvious prediction is that one will be more
likely to repeat inserting money in the useful-information
edition than in the basic edition. This reflects the instrumental value of feedback. Feedback carries valuable information about the quality of one’s behavior and can help
one learn and improve. Therefore, people like to receive
feedback (Baumeister 1998; Levy et al. 1995; London and
Smither 2002; Smither, London, and Reilly 2005) and are
motivated to engage in activities that provide feedback
(Ammons 1956; Bangert-Drowns et al. 1991; Finkelstein
and Fishbach 2012; Hattie and Timperley 2007; Kuhnen
and Tymula 2012; Locke, Cartledge, and Koeppel 1968;
Shute 2008; Sweetser and Wyeth 2005).
The present research is not about the preceding predictions regarding positive valence or useful information, that
is, not about the comparison between the basic edition and
the positive-valence edition of the tip jar, or the comparison between the basic edition and the useful-information
edition of the tip jar.
Rather, the present research is about the comparison between the basic edition and the mere-reaction edition of the
tip jar. The existing literature suggests one is equally or less
likely to repeat inserting money in the mere-reaction edition
than in the basic edition. Specifically, in terms of useful information, the two editions are the same. Even though the
mere-reaction edition emits a sound (e.g., an annoying beep)
when one tips, the sound provides no additional information
than what one would know anyway, that is, one can easily
see and tell that she has inserted money in the jar, regardless
of whether the sound exists or not. In terms of valence, operant conditioning predicts one to be less likely to repeat tipping in the mere-reaction edition than in the basic edition
because the sound (e.g., the annoying beep) is pretested to be
negative, hence a punishment. But we predict the opposite.
THEORY AND HYPOTHESES
Reaction
45
50
Will a person be more likely to repeat a behavior if she
experiences a reaction to the behavior than if not? This
question concerns a fundamental issue in psychology—the
relationship between stimulus (reaction) and response
(behavior). This question also carries potential practical
implications for designing products and programs such
as tip jars, donation boxes, exercise equipment, human–
computer interfaces, and loyalty programs.
As the tip jar example illustrates, prior research predicts
that people like to seek reactions if the reactions are a priori positive or provide useful information for learning and
improvement. We propose that even a priori nonpositive
and noninformative reactions can reinforce behaviors.
The question, then, is why. Our theory consists of two
components. The first component is that, in many situations,
reactions do carry useful information, such as how well one
has accomplished the objective one intends to accomplish,
and about how big an influence one has exerted on the domain one wants to influence. For example, the score a student receives after an exam informs him how well he has
performed in the class. The applause or boos a musician receives after a concert tell her how well she has performed
on the stage. Such reactions convey useful information—information that can help one learn from the past and improve
in the future. Consequently, people have learned to seek reactions and engage in behaviors that generate reactions.
The second component of our theory draws on the
notion of overgeneralization. Habits and heuristics are
learned and internalized in situations in which they are
functional, but they can continue to operate even in situations in which they serve no particular functions (Amir and
Ariely 2007; Arkes and Ayton 1999; Baron 2000). For example, the waste-not heuristic is learned in situations in
which wasting would deprive one of important resources
and worsen one’s future well-being, but the heuristic continues to influence one’s decisions even in situations in
which wasting means ignoring a sunk cost and would not
affect one’s future well-being (Arkes and Ayton 1999).
Likewise, we suggest that the tendency to seek reactions is
so well learned and internalized that it applies not only to
situations in which reactions carry useful information, but
also to situations in which they do not. That is, people like
to repeat behaviors that generate reactions even if the reactions are nonuseful. The mere presence of a reaction entails
a positive utility and makes the corresponding behavior engaging. This proposition concurs with casual observations
that many individuals are fond of squeezing bubble wrap
that produces popping sounds and pressing the button of a
ballpoint pen that generates a clicking sound, even though
the individuals find the sounds neither pleasing nor useful.
In summary, we propose the following as the primary
hypothesis of our research:
H1 (Reaction): A person will be more likely to repeat a behavior if the behavior is followed by a reaction than if it is
not, even if the reaction is a priori nonpositive and carries
little or no useful information.
55
60
65
70
75
80
85
90
95
Contingency and Valence Change
In addition to the primary hypothesis about reaction, we
propose two secondary hypotheses: one about contingency
and one about valence change. According to our theory,
the reason why a priori nonpositive and nonuseful stimuli
can reinforce a behavior is that they are reactions. For a
stimulus to be a reaction, the occurrence of the stimulus
must be contingent on the behavior. A stimulus is
100
105
HSEE ET AL.
5
10
contingent on a behavior if it is presented immediately following the behavior, and it is less or not contingent on the
behavior if it is presented either following a delay or in a
fixed or random interval independent of whether or when
the behavior occurs (Ferster and Skinner 1957; Skinner
1969). The contingency of a stimulus on a behavior is critical for the stimulus to be perceived as a reaction to the behavior and hence to possess a reinforcing value. Thus we
submit the following prediction as a qualification to our
primary hypothesis:
H2 (Contingency): The reaction effect postulated in H1
will be weaker if the stimulus is not contingent on the behavior than if it is.
15
20
25
30
35
40
45
The effect postulated in hypothesis 2 differs from the
contingency effect in the behavioral learning literature, according to which, if the stimulus is a priori negative (a punishment), it will be more effective in weakening a behavior
when it is contingent on the behavior than when not
(Ferster and Skinner 1957; Skinner 1969; Sweetser and
Wyeth 2005). Our theory predicts the opposite: even if the
stimulus is a priori negative, it can still be more effective
in reinforcing a behavior when it is contingent on the behavior than when it is not. Our reason is that contingency
leads the person to perceive the stimulus as a reaction.
Another secondary hypothesis we propose concerns the
valence of the reaction-serving stimulus. Our theory posits
that reactions are appealing and engaging beyond their a
priori valence and information values. It follows that if a
stimulus plays the role of a reaction to one’s action, the
person will come to like it and will perceive it to be positive even if the stimulus is a priori nonpositive. In other
words, the valence of a stimulus can change when the stimulus serves as a reaction. This is our next hypothesis:
H3 (Valence Change): An a priori nonpositive stimulus
will become more positive by serving as a reaction to one’s
behavior.
As should be evident from the preceding analysis, the
contingency and the valence-change hypotheses are not independent of the primary hypothesis about reaction; rather,
they are supplementary to the primary hypothesis. The contingency hypothesis explains when a stimulus becomes a
reaction and hence when it carries a reinforcing value. The
valence-change hypothesis differentiates the reaction effect
from standard operant conditioning, suggesting that a stimulus acquires a positive valence because it is a reaction
rather than that a stimulus carries a reinforcing value because it is a priori positive.
Related Findings in the Existing Literature
50
Several existing bodies of literature have inspired the
current work. One is on behavioral learning. Most studies
in this area have observed the conventional operant
3
conditioning effects that a priori positive stimuli reinforce
behaviors and a priori negative stimuli weaken behaviors,
whereas a priori neutral stimuli have no effect (e.g.,
Nuthmann 1957; Taffel 1955). But there are notable exceptions. One set of studies on human learning found that
neutral reactions, such as a tapping response by the experimenter, could reinforce participants’ spontaneously occurring behaviors, such as smiling or touching the nose
(Verplanck 1956). This finding might reflect a reaction effect. However, as Verplanck noted, the participants in
those studies were typically informed in advance about the
meaning of the responses, for example, were told in advance that they would receive a point every time the experimenter tapped, and if they accrued enough points they
would be rewarded. Therefore, it was unclear whether the
result was due to the mere presence of a reaction or due to
its association with the points and the reward. Similarly,
prior research on token economics has demonstrated that
neutral tokens, such as chips and points, can serve as reinforcers (e.g., Boerke and Reitman 2011; Hsee et al. 2003;
Kazdin 1982; Nunes and Drèze 2006). However, even
though the tokens are a priori neutral, they can be
redeemed for rewards such as cash and candies, hence associated with positive outcomes. In the context of animal
learning, Muenzinger (1934) found that even negative
stimuli, such as electric shocks, could reinforce the searching behavior of hungry rats for food. Again, it was unclear
whether the reinforcing value of the electric shocks
stemmed from their role as reactions or their association
with food acquisition. In our studies, we do not associate
reactions with any external rewards.
More pertinent to the current research are studies in the
behavioral learning literature showing that even nonpositive stimuli not associated with external rewards can reinforce behaviors. In one such study, Greenspoon (1954)
asked human participants to utter nouns in singular or plural forms. One group of participants would see the flash of
a light if they uttered a plural noun, and another group
would see such a signal if they uttered a singular noun.
Over time, participants who saw the light upon uttering
plural nouns uttered more plural nouns, and participants
who saw the light upon uttering singular nouns uttered
more singular nouns (see Sidowski 1954 for a similar finding). In another study, Tolman, Hall, and Bretnall (1932)
asked human participants to solve a maze puzzle. Among
different groups in the study, one group would hear a bell
tone and receive an electric shock every time they made a
correct move, and another group would hear a bell tone
and receive an electric shock every time they made a
wrong move. The former group learned faster to solve
the puzzle than the latter group. In animal learning,
Kling, Horowitz, and Delhagen (1956) found that rats confined in a cage would press a bar more frequently if the
bar-pressing action resulted in a change in ambient lighting
than if it did not (see Girdner 1953; Hurwitz 1956; and
55
60
65
70
75
80
85
90
95
100
105
4
5
10
15
20
25
30
35
40
45
50
55
Roberts, Marx, and Collier 1958 for similar findings).
These studies suggested that mere reactions can be reinforcing. Nevertheless, the stimuli in the studies involving
humans may have conveyed useful information. For example, the light in Greenspoon’s study may have suggested
whether the experimenter wanted the participant to utter
singular or plural nouns (even if the participant were unable to verbalize the relationship), and the bell and the
electric shock in study by Tolman et al. may have suggested whether the participant’s move in the maze was correct or wrong. In the current research, we try to minimize
the information value of the reactions.
Another body of literature that has inspired our work is
on control. People like to have control over their environment and like to exercise control when they can (e.g.,
Bandura 1997; Burger 1992; Deci and Ryan 2000;
Hamerman and Johar 2013; Inesi et al. 2011; Rodin and
Langer 1976; Rucker and Galinsky 2008; Ryan, Rigby,
and Przybylski 2006; White 1959; Whitson and Galinsky
2008). The notion of control can potentially explain the
reaction effect, as well as valence change and contingency: a reaction is a signal that one has successfully exerted control and thereby makes the person feel good
about the signal; contingency heightens the sense of control and hence increases the effect of the reaction. At first
glance, this control-based explanation is different from
our explanation, which attributes the reaction effect to an
overgeneralized tendency to seek useful information. But
at a deeper level, we believe the two explanations are connected. Reactions often signal one’s ability to control
what is important, such as resources in the environment,
and can help one learn and improve. Thus to seek reactions is to seek information about one’s power in controlling something important. However, people overlearn and
overgeneralize this tendency, and they will seek reactions
even when the reactions do not provide information about
one’s ability to control anything important but rather only
signal one’s influence over something trivial, such as a
sound.
Our research also builds on the existing literature showing that humans, as well as many other animals (such as
cats), are curious about their environments and voluntarily
explore (e.g., Berlyne 1966; Groos 1901; Montgomery
1952; Piaget 1952). In many circumstances, exploration is
useful because it can help individuals better adapt to their
environments or make better decisions (e.g., Dervin 1998;
Piaget 1969; Ratchford, Talukdar, and Lee 2007).
However, when exposed to a gap in information, individuals will feel curious and seek the missing information,
even if the information serves no other function than the
resolution of the curiosity (e.g., Asch, Patton and Hershey
1990; Bastardi and Shafir 1998; Golman and Loewenstein
2012; Loewenstein 1994). Whereas the current research is
not about exploration or curiosity per se, it corroborates the
exiting literature in the area by highlighting people’s
JOURNAL OF CONSUMER RESEARCH
tendency to seek information, even when the information
is apparently useless.
OVERVIEW OF STUDIES
We next report eight experimental studies including a
field experiment. All of the studies tested our primary
hypothesis regarding reaction. Some of the studies also
tested the secondary hypotheses concerning contingency or
valence change and sought to shed light on the mechanisms
underlying the reaction effect.
We determined the a priori valence of the stimulus used
in each study in a pretest. Participants in the pretest experienced the stimulus in the same environment as that of the
main study and rated the stimulus on a negative–positive
scale. For example, in study 1, the stimulus was a naturally
occurring sound when a metal object was hit; participants
in the pretest listened to the same sound from the same
object at the same distance as participants in the main
study, and they rated it on a 7-point scale anchored by
“very negative” (left) and “very positive” (right). We describe the sample size and the scale used in each pretest
when reporting the corresponding main study.
In most of our studies, we used only a priori mildly
negative stimuli, namely, stimuli pretested to be moderately but significantly below the midpoint of the negative–
positive scale. Why did we use a priori mildly negative
stimuli, and not a priori positive stimuli or a priori very
negative stimuli? That is because the latter two types of
stimuli would not allow us to distinguish our theory from
operant conditioning. If the stimulus is a priori positive,
both our theory and operant conditioning predict a reinforcing effect. If the stimulus is a priori very negative (e.g.,
a painful electric shock), operant conditioning will predict
a strong weakening effect, so strong that it will eclipse any
reinforcing effect of reaction. In contrast, a priori mildly
negative stimuli enable us to distinguish our theory from
operant conditioning: our theory predicts a reinforcing effect, whereas operant conditioning predicts a weakening
effect. To reiterate, we are not downplaying operant conditioning; rather, we wish to show that beyond it, the mere
presence of a reaction also matters.
In real life, marketers typically use a priori positive stimuli rather than a priori negative stimuli to reinforce consumer behaviors. It is for theory-testing purposes that we
used a priori negative stimuli in our studies. For external
validity, we also included a priori positive stimuli in two of
our studies.
60
65
70
75
80
85
90
95
100
STUDY 1: BEAN THROWING
Study 1 sought to demonstrate the reaction effect in
a minimalistic setting. The behavior was a light physical task, and the reaction was a naturally occurring sound.
105
HSEE ET AL.
5
TABLE 1
RESULTS OF ALL THE STUDIES
Study
Dependent variables
No reaction
Contingent stimulus (reaction)
Noncontingent stimulus
Study 1
Study 2
Study 3
Behavior (beans thrown)
Behavior (payment)
Behavior (times jumped)
Willingness to continue
Engagement
Behavior (passwords entered)
Perception of reaction
Feeling toward stimulus
Behavior (messages sent)
Behavior (messages sent)
Behavior (coins donated)
36.56 (17.20)
3.80 (3.20)
12.07 (6.40)
2.80 (1.16)
2.77 (1.04)
32.48 (19.31)
NA
Pretest .54 (1.00)
7.06 (3.82)
10.61 (7.76)
2.76 (1.71)
49.92 (26.46)
6.92 (2.92)
18.11 (12.74)
3.46 (1.11)
3.64 (.99)
43.34 (20.70)
3.23 (1.03)
.23 (.78)
Positive 12.89 (10.05) Negative 10.21 (5.60)
Positive 15.88 (9.77) Negative 13.52 (8.59)
3.92 (2.69)
NA
NA
12.00 (6.08)
2.77 (.77)
3.07 (.83)
33.03 (20.55)
1.49 (.80)
.46 (.92)
NA
NA
NA
Study 4
Study 5
Study 6
Study 7
Study 8
NOTE.—Means are outside the parentheses; standard deviations are inside. In all the cases, values were greater in the contingent-stimulus (reaction) condition
than in both the no-reaction condition and the non–contingent-stimulus condition (wherever applicable).
Method
5
10
15
20
25
Fifty students (12 females, Mage ¼ 22.86 years, standard
deviation [SD] ¼ 2.29 years) recruited from a large public
university participated in the study for a fixed nominal payment. They were run individually in a private room. In
front of each participant was a bucket of large soybeans,
and 5 feet in front of the participant was a target—a white
disk 1 foot in diameter. Participants were told that the
study was about a light physical exercise; they needed to
spend 3 minutes throwing the beans, one at a time, at the
target, and then answer a questionnaire about the activity.
Participants were further told that they could repeat the
task as few or as many times as they wished during the 3
minute period, that the activity was not a contest, and that
their payment would not depend on their performance.
During the period, a research assistant stood nearby and informed the participant when the time was up.
Participants were assigned to one of two conditions: no
sound (i.e., no reaction) and sound (i.e., reaction). The only
difference between the conditions was in the material of
which the target was made. In the no-sound condition, the
target was made of sponge, so that when a bean hit it, it
was largely mute. In the sound condition, the target was
made of metal, so that when a bean hit it, it would emit a
sound. In a pretest (N ¼ 23), the sound was rated below the
midpoint of a 7-point scale anchored by “Very negative”
(left) and “Very positive” (right) (M ¼ 1.39 when coding
the midpoint as 0, SD ¼ .72; t(22) ¼ 9.24, p < .001).
Results and Discussion
30
35
As Table 1 summarizes, participants in the sound condition repeated the bean-throwing activity more times
(M ¼ 49.92, SD ¼ 26.46) than those in the no-sound condition (M ¼ 36.56, SD ¼ 17.20; F(1, 48) ¼ 2.12, p < .05).
The two conditions did not differ in hit rate (72% vs. 70%,
not significant [NS]).
This result illustrates the reaction effect postulated in hypothesis 1. Note that the sound in the study was pretested
to be negative, and therefore it was not a reward in the traditional sense of operant conditioning. Moreover, the
sound carried no useful information; that is, it conveyed no
more information about one’s performance than what one
would know anyway: the target was only a few feet from
the participant and regardless of whether there was a
sound, the participant could easily tell whether the target
was hit.
The result of study 1 could not be attributed to demand
(guessing of the experimenter’s intention) because the only
reaction in the sound condition was that of the bean hitting
the metal target, which occurred naturally and would not
suggest the intention of the experimenter. The result could
not be explained in terms of goal progress or goal gradient
either (e.g., Nunes and Drèze 2006; Zhang and Huang
2010). Participants may have set a goal of hitting the target
a certain number of times, but that could not explain the
difference between the sound and the no-sound conditions.
This study provides initial evidence that reactions are appealing and reinforcing beyond their valence and information value.
Since reactions are appealing and reinforcing, readers
may ask why individuals do not repeatedly clap their hands
or bang their heads against the wall to make noise. The reason is simple: doing these things incurs costs, such as effort
and pain. Our theory makes no prediction about the absolute likelihood that a person would engage in or repeat a reaction-generating activity. Rather, it predicts that paribus
ceteris, a person is more likely to repeat an activity if it
generates reactions than if it does not.
40
45
50
55
60
65
STUDY 2: PAY WHAT YOU WANT
Study 2 was a replication of study 1 in a more consumer-relevant “pay-what-you-want” context. Pay what
70
6
5
10
15
20
25
30
35
40
you want is a marketing strategy whereby consumers voluntarily pay any amount for a product as they wish (e.g.,
Gneezy et al. 2010; Kim, Natter, and Spann 2009). The
purpose of this study was not to compare this strategy with
other strategies. Rather, it was to compare two editions of a
payment box—one with and one without a sound (like the
mere-reaction tip jar and the basic tip jar in the opening example), and see which was more effective in soliciting
contributions.
Method
This experiment took place after the participants—62
students (45 females, Mage ¼ 20.58 years, SD ¼ 3.44 years)
recruited from a large private university—had completed
another unrelated experiment. Specifically, after each participant had completed the first experiment, a research assistant took him to a small room and paid him the
promised $2.00 for that experiment. The payment consisted
of a $1.00 bill and 10 dimes. In addition, the research assistant gave the participant a Uni-ball Roller Grip Roller Ball
pen (retail price $1.50), said that he could keep the pen,
and suggested that he pay what he wanted for the pen and
drop the money in a payment box. The research assistant
then left the room. The payment box sat on a table in the
room and had a narrow opening at the top, through which
one could insert coins. Attaching to the box was a sign,
which said in large font, “Pay What You Want/Use Dimes
Only!/Drop One Dime at a Time!”
The payment box had two editions: sound and no sound.
The two editions were identical except that inside the
sound edition we placed a piece of metal at the bottom so
that when a coin was dropped, it would emit a noise,
whereas inside the no-sound edition, we placed a layer of
cloth so that it would not emit any sounds. In a pretest
(N ¼ 24), the metal noise was rated to be significantly below the midpoint of a 7-point scale anchored by “Very
negative” (left) and “Very positive” (right) (M ¼ .33
when coding the midpoint as 0, SD ¼ .64; t(23) ¼ 2.56,
p < .05). During the study, the research assistant randomly
alternated between the sound edition and the no-sound edition between participants.
Results and Discussion
45
50
Among the participants in this study, most (79%) paid at
least a dime for the pen, and the percentage of participants
who paid did not differ significantly between the two conditions (83% vs. 76%, NS). Importantly, among the participants who paid, those in the sound condition paid
significantly more (M ¼ 6.92, SD ¼ 2.92) than those in
the
no-sound
condition
(M ¼ 3.80,
SD ¼ 3.20;
F(1, 47) ¼ 12.65, p < .01). (Even if we included those who
did not pay anything, the effect was still significant;
F(1, 60) ¼ 4.81, p < .05.) Study 2 replicated the reaction
JOURNAL OF CONSUMER RESEARCH
effect observed in study 1: the presence of a naturally occurring noise reinforced the coin-inserting behavior, raising payment by over 80%.
STUDY 3: JUMPING
55
Study 3 involved an exhausting physical task—jumping.
Unlike study 1 and study 2 that included only two conditions (no sound and sound), study 3 included three conditions: no sound, contingent sound, and noncontingent
sound. This design enabled us to test both the reaction
hypothesis (hypothesis 1) and the contingency hypothesis
(hypothesis 2). In addition to overt behavior, study 3
also measured willingness to continue and feeling of
engagement.
60
Method
65
Ninety male students (Mage ¼ 21.19 years, SD ¼ 1.03
years) recruited from a large public university participated
in the experiment for a fixed nominal payment. (We used
only male participants because men and women have different physical strengths, and the use of participants of the
same sex reduces the variance.) The participants were run
individually facing a wall in a private room. A large touchpad hung on the wall. We adjusted the touchpad for participants of different heights, so that everybody had to jump a
bit to touch it. Participants were told that their task was to
perform a physical exercise—to jump repeatedly to touch
the touchpad and then answer a few questions afterward.
Each participant was given 2 minutes to do the exercise
and told that he should jump at least five times and after
that he could choose to jump more times as he pleased.
During the period, a research assistant stood nearby and
would tell the participant when the time was up.
Participants were randomly assigned to one of three conditions: no sound, contingent sound, and noncontingent
sound. In the no-sound condition, no sound was presented.
In the contingent-sound condition, as the participant
touched the touchpad, it immediately emitted a brief (0.8 s)
sound. In a pretest (N ¼ 22), the sound was rated significantly below the midpoint of a 9-point scale anchored by
“Very negative” (left) and “Very positive” (right)
(M ¼ .91 when coding the midpoint as 0, SD ¼ 1.48;
t(21) ¼ 2.89, p < .01). The non–contingent-sound condition was identical to the contingent-sound condition except
that the sound was not contingent on jumping. Instead, the
touchpad emitted the sound a total of 15 times over the
2-minute period, with random intervals in between. We
chose 15 times because in a pretest, we found that participants jumped 15 times on average in the same duration of
time.
At the end of the period, participants were asked whether
they were willing to continue if they could and whether
they found the physical activity engaging. They answered
70
75
80
85
90
95
100
HSEE ET AL.
5
each question on a 5-point scale ranging from 1 (Not at all)
to 5 (Very much). Prior to analysis, we excluded two participants: one jumped fewer than five times (the required
minimum number), and the other jumped more than three
SDs from the mean. Including them did not change the pattern of the results.
Results and Discussion
10
15
20
25
30
35
40
45
50
Behavior. The main dependent variable was the number
of times participants jumped. Table 1 presents the results.
A one-way analysis of variance (ANOVA) across the three
conditions revealed a significant effect (F(2, 85) ¼ 4.51,
p < .05). To test for the reaction effect, we conducted a
planned comparison between the contingent-sound and the
no-sound conditions and found the predicted effect—that
participants in the contingent-sound condition jumped
more times (M ¼ 18.11, SD ¼ 12.74) than those in the nosound condition (M ¼ 12.07, SD ¼ 6.40; F(1, 85) ¼ 6.76,
p < .05). To test for contingency, we conducted a planned
comparison between the contingent-sound and the non–
contingent-sound conditions and also found the predicted
effect—that participants in the contingent-sound condition
jumped more than those in the non–contingent-sound condition (M ¼ 12.00, SD ¼ 6.08; F(1, 85) ¼ 6.91, p < .05).
There was no significant difference between the no-sound
and the non–contingent-sound conditions (F(1, 85) ¼ .001,
p > .10). In summary, study 3 replicated the reaction effect
and provided evidence for the importance of contingency.
Willingness to Continue. At the end of the designated
period, we asked participants whether they were willing to
continue. Normatively, participants in the contingentsound condition should have been the least willing to continue the task because they had already jumped the most
number of times and must have been the most exhausted.
Yet the result was the opposite of this normative prediction
and was consistent with the jump results: participants in
the contingent-sound condition were more willing to continue (M ¼ 3.46, SD ¼ 1.11) than participants in both the
no-sound condition (M ¼ 2.80, SD ¼ 1.16; F(1, 85) ¼ 6.10,
p < .05) and the non–contingent-sound condition
(M ¼ 2.77, SD ¼ .77; F(1, 85) ¼ 6.72, p < .05). Providing a
contingent stimulus seemed to not only prompt participants
to jump more in the given time period but also to entice
them to extend the time period.
Feeling of Engagement. In addition to the behavioral
variables, we also asked participants to rate the engagingness of the activity. Participants in the contingent-sound
condition found the activity more engaging (M ¼ 3.64,
SD ¼ .99) than both participants in the no-sound condition
(M ¼ 2.77, SD ¼ 1.04; F(1, 85) ¼ 12.17, p < .01) and participants in the non–contingent-sound condition (M ¼ 3.07,
SD ¼ .83; F(1, 85) ¼ 5.26, p < .05). In line with the
7
behavioral results, the presence of reactions made the activity feel engaging.
STUDY 4: PASSWORD ENTRY
Study 4 sought to replicate the effects of reaction and
contingency in a cognitive (rather than physical) task and
using a visual (rather than audio) stimulus. The study also
examined the perception of reaction, namely, the extent to
which participants perceived the stimulus as a reaction to
their action.
55
60
Method
A total of 118 adults (33 women, Mage ¼ 29.91 years,
SD ¼ 8.39 years) recruited from Mechanical Turk (MTurk)
participated in the study for a fixed nominal payment.
Participants were told their task was to type the password
“lockit12” and enter it by pressing the ENTER key. They
were told they had to spend 2 minutes on the task, and during the period they could repeat the task as few or as many
times as they wished. Once the study began, participants
could not exit for the first 2 minutes; after that, an exit button appeared and participants could click it to exit.
Participants were randomly assigned to one of three conditions: no circle, contingent circle, and noncontingent circle. In the no-circle condition, no image appeared on the
screen during the study. In the contingent-circle condition,
a solid brown color circle appeared at the center of the
screen every time the participant entered the password and
then it rapidly shrank until it disappeared; the total duration
was about 0.3 s. In a pretest (N ¼ 41), the brown circle was
rated to be significantly below the midpoint of a 5-point
scale anchored by “Very negative” (left) and “Very positive” (right) (M ¼ .54 when coding the midpoint as 0,
SD ¼ 1.00; t(40) ¼ 3.43, p < .01). In the non–contingentcircle condition, the brown circle also appeared, but its appearance did not depend on the entering of the password; it
appeared at a fixed frequency of either once every 1 s or
once every 9 s. The 1 s and the 9 s versions did not yield
significantly different results and therefore were combined
in subsequent analyses.
Before they started the task, participants in all the
conditions were asked to focus their attention on the screen
during the task. In the contingent-stimulus and the non–
contingent-stimulus conditions, participants were also told
that in order to focus their attention on the screen, a brown
circle would flash on the screen from time to time. In the
contingent-stimulus condition, participants were told that
how frequently the circle flashed on the screen would depend on how frequently they entered the password and that
it would flash once every time they entered the password.
In the non–contingent-stimulus condition, participants
were told that how frequently the circle flashed on the
65
70
75
80
85
90
95
100
8
5
10
screen was predetermined and would not depend on how
frequently they entered the password.
In all the conditions, participants could see what they
were typing, and only if they had typed correctly were they
able to enter the password and clear it from the screen.
Therefore, the circle provided no additional information
about whether they had typed the password correctly.
To assess participants’ perception of reaction at the end
of the study, we asked participants in the contingent and
non–contingent-stimulus conditions to rate the extent to
which the occurrence of the brown circle felt like a reaction to their entering the password on a 4-point scale ranging from 1 (Not at all) to 4 (Very much).
Results and Discussion
15
20
25
30
35
40
45
Behavior. The main dependent variable was the number
of times participants entered the password. Table 1 presents the results. A one-way ANOVA across the three conditions revealed a significant effect (F(2, 115) ¼ 3.57,
p < .05). A planned comparison between the contingentcircle and the no-circle conditions replicated the reaction
effect: participants in the contingent-circle condition entered the password significantly more times (M ¼ 43.34,
SD ¼ 20.70) than participants in the no-circle condition
(M ¼ 32.48, SD ¼ 19.31; F(1, 115) ¼ 5.79, p < .05). A
planned comparison between the contingent-circle and the
non–contingent-circle conditions replicated the contingency effect: those in the contingent-circle condition
entered the password more times than those in the non–
contingent-circle condition (M ¼ 33.03, SD ¼ 20.55;
F(1, 115) ¼ 4.97, p < .05). There was no difference between the non–contingent-circle and the no-circle conditions (F(1, 115) ¼ .01, p > .10).
Perception of Reaction. As expected, participants in
the contingent-circle condition perceived the occurrence
of the brown circle to be significantly more like a reaction
to their password-entry behavior (M ¼ 3.23, SD ¼ 1.03)
than did participants in the non–contingent-circle
(M ¼ 1.49, SD ¼ .80; t(70) ¼ 8.02, p < .01). A mediation
analysis (Preacher and Hayes 2008) found this reactionperception variable fully mediated the effect of contingency on the behavior, with the indirect effect (9.99) for
the overall model differing from zero at the 95% confidence interval, .17–18.60. It appears that perceiving a
stimulus to be a reaction is critical for the stimulus to be
reinforcing.
STUDY 5: PASSWORD ENTRY (VALENCE
CHANGE)
50
Study 5 tested the valence-change hypothesis (hypothesis 3) that an a priori negative stimulus can become positive by serving as a reaction.
JOURNAL OF CONSUMER RESEARCH
Method
A total of 104 four adults (33 women, Mage ¼ 30.89 years,
SD ¼ 9.52 years) recruited from MTurk participated in the
study for a fixed nominal payment. The study used the same
stimulus (the brown circle) and the same procedure as study
4 except for the following. Study 5 included only one
dependent variable—feelings toward the stimulus—and included only two between-participants conditions: contingent
circle and noncontingent circle. (As in study 4, the non–
contingent-circle condition consisted of two subconditions,
in one of which the circle appeared every 1 s and in the other
the circle appeared every 9 s. These subconditions did not
yield significantly different results and were combined
in further analyses.) In both the contingent and the non–
contingent-circle conditions, we required participants to enter the password 30 times and then asked them to rate their
feelings toward the circle on the same scale as in the pretest.
55
60
65
Results and Discussion
As Table 1 summarizes, participants in the contingentcircle condition rated the stimulus (the circle) more positively (M ¼ .23, SD ¼ .78) than those in the pretest
(M ¼ .54, SD ¼ 1.00; t(91) ¼ 4.15, p < .001). In fact, the
circle, which was rated below the midpoint of the positivenegative scale in the pretest, was now rated significantly
above the midpoint (t(51) ¼ 2.13, p < .05). Moreover, participants in the non–contingent-circle condition did not rate
the circle differently (M ¼ .46, SD ¼ .92) from those in
the pretest (t(91) ¼ .38, p > .10). Importantly, however,
participants in the contingent-circle condition rated the
circle more positively than did participants in the non–
contingent-circle condition (t(102) ¼ 4.14, p < .001).
Consistent with the valence-change hypothesis (hypothesis
3), the circle felt more positive in the main study than in
the pretest, and it did so only if the circle acted as a reaction (i.e., was contingent on the behavior) and not if the circle merely appeared (i.e., was not contingent). The latter
result ruled out mere exposure (Fang, Singh, and
Ahluwalia 2007; Zajonc 1968) as an alternative account.
This study offered initial evidence that an a priori negative stimulus can become positive by serving as a reaction.
In our opinion, the valence change does not reduce the reaction effect to a mere operant-conditioning effect in its traditional sense, according to which a stimulus must be a priori
positive to be reinforcing. Rather, the valence change highlights the reaction effect, both effects corroborating our
proposition that people like to see reactions to their actions.
70
75
80
85
90
95
STUDY 6: TEXT MESSAGING (BETWEEN
PARTICIPANTS)
Thus far we have used only a priori negative stimuli. To
test the generality of the reaction effect and show that it
100
HSEE ET AL.
5
will arise with both a priori positive and negative stimuli,
we conducted study 6. The study included three conditions:
positive reaction, negative reaction, and no reaction (control). For simplicity, the stimuli were always contingent on
the behavior.
Method
10
15
20
25
30
35
40
45
50
A total of 130 adults (75 women, Mage ¼ 38.55 years,
SD ¼ 12.35 years) recruited from MTurk participated in the
study for a fixed nominal payment. Participants were told
that the purpose of the study was to test user experience of
the text message interface, and that their task was to type
and send short messages on the computer and answer questions afterward. Participants were told that they could type
any messages that they wished; they were encouraged to repeat the task as many times as possible but were told that
their payment would be the same. During the study, participants would see a textbox at the center of the screen and a
SEND button beneath it. They would type their message in
the textbox and press the SEND button afterward.
Participants could see what they typed in the textbox, and
the message would disappear once they pressed the SEND
button. These features were common to all participants.
Participants were randomly assigned to one of three conditions: positive reaction, negative reaction, and no reaction. In the negative-reaction condition, the image of a
cockroach appeared for about 0.4 s every time participants
sent a message. In the positive-reaction condition, the image of a butterfly appeared for the same duration every
time participants sent a message. In the no-reaction condition, no image appeared. In a pretest (N ¼ 30), the cockroach image was rated to be significantly below the
midpoint of a 5-point scale anchored by “Very negative”
(left) and “Very positive” (right) (M ¼ 1.03 when coding
the midpoint as 0, SD ¼ .72; t(29) ¼ 7.88, p < .001), and
the butterfly image was rated to be significantly above the
midpoint (M ¼ 1.40, SD ¼ .62; t(29) ¼ 12.34, p < .001).
Before they started the task, participants in each condition
were asked to type and send a message in the textbox so
that they knew what their textbox was like.
We did not impose a time constraint in this study; participants could exit the task at any time. After they exited,
participants completed a 20-item Desirability for Control
Scale (Burger and Cooper 1979) that assessed individual
differences in their desire to control events in their lives.
Prior research suggests that individuals who lack power or
have a high desire for control behave differently than
others (e.g., Hamerman and Johar 2013; Inesi et al. 2011;
Rucker and Galinsky 2008; Whitson and Galinsky 2008).
We were curious whether such individuals also respond
more to reactions.
Prior to analysis, we excluded two participants: one
in the no-reaction condition and the other in the positivereaction condition, both of whom typed more than three
9
SDs from the mean. Including them did not change the pattern of the results.
55
Results and Discussion
Behavior. The main dependent variable was the number
of times participants sent messages. Table 1 presents the
results. A one-way ANOVA across the three conditions revealed a significant effect (F(2, 125) ¼ 7.99, p ¼ .001).
Further analyses found the following: participants in both
the positive-reaction condition (M ¼ 12.89, SD ¼ 10.05)
and the negative-reaction condition (M ¼ 10.21,
SD ¼ 5.60) repeated the activity more times than participants in the no-reaction condition (M ¼ 7.06, SD ¼ 3.82;
F(1, 125) ¼ 15.72, p < .001, and F(1, 125) ¼ 4.97, p < .05,
respectively), indicating that relative to no reaction, both
positive and negative reactions were reinforcing.
Moreover, participants in the positive-reaction condition
repeated the task somewhat more times than participants in
the negative-reaction condition (F(1, 125) ¼ 3.17,
p ¼ .077), which likely reflected operant conditioning.
Desire for Control. We found no correlation between
participants’ behaviors (messages sent) and their desirefor-control scores, and this was true regardless of whether
we ran the correlation analysis across all the three reaction
conditions or within each condition (all r values < .17; all
p values > .29).
60
65
70
75
STUDY 7: TEXT MESSAGING (WITHIN
PARTICIPANTS)
80
All of the studies reported so far followed a betweenparticipants design. To test the robustness of our finding,
study 7 adopted a within-participants design, including
three within-participants conditions: positive reaction, negative reaction, and no reaction (control).
85
Method
A total of 87 adults (56 women, Mage ¼ 34.45 years,
SD ¼ 11.73 years) recruited from MTurk participated in
the study for a fixed nominal payment. The procedure of
this study was similar to that of study 6 except for the following. The study presented three (rather than one) textboxes : positive reaction, negative reaction, and no
reaction. The three textboxes were displayed horizontally
in the lower part of the screen, and their positions (left,
middle, or right) were randomized across participants.
Above each textbox was a neutral-looking face (a circle
with three dots representing the eyes and the mouth), and
below each textbox was a SEND button.
Participants were told that they must type and send the
message “hello” 40 times during the study, that they could
use any of the textboxes, and that the computer would keep
track and tell them when they had reached the required
90
95
100
10
5
10
15
20
25
30
number. As in study 6, participants could see what they
typed and could see the message disappear once they hit
the SEND button. These features were common to all the
conditions.
The manipulation lay in the face image above each textbox. If a participant entered the message in the positivereaction textbox, the neutral face above it would change to
a positive-looking face (with three downward triangles representing the eyes and the mouth) for 0.4 s and then return
to the original neutral condition. If a participant entered the
message in the negative-reaction textbox, the neutral face
above it would change to a negative-looking face (with
three upward triangles representing the eyes and the
mouth) for 0.4 s and then return to the original neutral
condition. If a participant entered the message in the noreaction textbox, the neutral face would not change. In a
pretest (N ¼ 30), the positive-looking face was rated to be
significantly above the midpoint of a 5-point scale anchored by “Very negative” (left) and “Very positive”
(right) (M ¼ .57 when treating the midpoint as 0,
SD ¼ 1.14; t(29) ¼ 2.73, p < .05), the negative-looking face
significantly below the midpoint (M ¼ .60, SD ¼ 1.25;
t(29) ¼ 2.63, p < .05), and the neutral face not significantly different from the midpoint (M ¼ .07, SD ¼ .58;
t(29) ¼ .63, p > .10). To simulate the context of the main
study, the stimuli were presented within participants in the
pretest, with their order randomized.
Before they started the task, participants tried each of
the three textboxes so they knew what each textbox was
like. After the study, participants completed the 20-item
Desirability of Control Scale as used in study 6.
Results and Discussion
35
40
45
50
Behavior. Recall that participants were required to type
and send “hello” 40 times. The key dependent variable was
the distribution of the 40 counts across the three textboxes.
Table 1 presents the results. Replicating the results of study
6, within-subjects t tests show that participants used both
the positive-reaction textbox (M ¼ 15.88, SD ¼ 9.77) and
the negative-reaction textbox (M ¼ 13.52, SD ¼ 8.59) more
times than they used the no-reaction textbox (M ¼ 10.61,
SD ¼ 7.76; t(96) ¼ 3.37, p ¼ .001, and t(96) ¼ 2.18,
p < .05, respectively), and they used the no-reaction textbox fewer times than by chance (t(96) ¼ 3.46, p ¼ .001).
There was no difference in usage between the positive- and
the negative-reaction textboxes (t(96) ¼ 1.39, p > .10).
Compared with the between-participants design, the
within-participants design had its pros and cons in addressing demand as an alternative explanation. The within-participants design made it transparent that two textboxes
generated reactions and one did not, and thereby may have
led participants to suspect that the experimenter wanted
them to use the textboxes with reactions. However, the
within-participants design also made it transparent that the
JOURNAL OF CONSUMER RESEARCH
reaction from one of the textboxes was negative and
thereby may have led participants to suspect that the experimenter wanted them to avoid it. The fact that we found
similar results in both the between-participants and the
within-participants designs rendered demand a nonparsimonious explanation.
Desire for Control. As in study 6, we found no correlation between participants’ desire-for-control scores and
their preference to use the textboxes with reactions over
the textbox without reactions (r ¼ .03, p ¼ .75).
STUDY 8: DONATION (FIELD
EXPERIMENT)
55
60
65
Study 8 was a field experiment that tested our primary
hypothesis (hypothesis 1) in a natural setting involving real
donation behaviors. Like study 1 and study 2, study 8 included two conditions: sound and no sound.
Method
We conducted the study in a coffee shop over a two-day
period. The shop was located near a large public university
in China, and most customers were students. The shop was
running a campaign to solicit donations to feed homeless
cats on campus. To advertise the campaign, a large sign
with the title “Cat Loving Day” was erected near the cashier; the sign read, “Please make a donation to show your
love for the stray cats on campus. . . . We promise we will
use all of the donations to buy food for the cats.” Next to
the sign and on the cashier counter sat an opaque donation
box. At the top of the box was a slot in which one could insert coins. Whenever a customer purchased something at
the cashier, a fundraiser for the event would alert him or
her to take a look at the sign and the donation box. To encourage customers to donate, the shop gave only 1-yuan
coins as change to customers. To donate money, the customer would insert coins one by one into the slot on top of
the donation box. If a customer was willing to donate but
had only paper bills, the fundraiser would exchange 1-yuan
coins for him or her.
The manipulation of the study lay in the donation box. A
small device beneath the slot of the donation box would
emit a brief (0.7 s) sound every time one inserted a coin.
The sound could be switched on or off, which allowed us
to have a sound condition and a no-sound condition. In a
pretest (N ¼ 27), the sound was rated to be significantly below the midpoint of a 5-point scale anchored by “Very negative” (left) and “Very positive” (right) (M ¼ .96 when
coding the midpoint as 0, SD ¼ .85; t(26) ¼ 5.86,
p < .01). The sound was of a moderate level so that the donor could hear it but other customers in the coffee shop
could not because the coffee shop had various ambient
sounds such as coffee grinding noise, conversations, and
70
75
80
85
90
95
100
HSEE ET AL.
5
10
15
20
background music. In both the sound and the no-sound
conditions, we placed a layer of cushioning material at the
bottom of the donation box to minimize the sound it would
make when a dropped coin hit the bottom. Furthermore, in
both conditions, we frequently emptied the donation box to
minimize the likelihood that a dropped coin would crash
against existing coins in the box, although it was impossible to totally eliminate this likelihood.
During each of the two days on which we conducted the
study, we alternated between the sound and no-sound conditions (by turning on or off the sound of the donation box)
every hour. During all times, the fundraiser counted the
number of customers who entered the shop, the number of
customers who donated, and the number of coins each donor contributed.
Three types of coins were in circulation when we conducted the study: 1 yuan (the most popular), 0.5 yuan, and
0.1 yuan. The fundraiser was unable to track the exact
types of coins each donor contributed, but we were able to
track the distributions of the different types of coins in
each condition as a whole, and found no significant difference between the two conditions (v2(2) ¼ .32, p ¼ .85).
Results and Discussion
25
30
35
40
45
50
During the periods in which we conducted the study, a
total of 392 customers visited the coffee shop: 203 in the
no-sound condition and 189 in the sound condition. Of
these customers, 41 (20.20%) in the no-sound condition
and 49 (25.92%) in the sound condition made donations,
and the two percentages were not significantly different
(v2(1) ¼ 1.13, p ¼ .29).
Importantly, among the customers who donated, those in
the sound condition donated more than those in the nosound condition. Overall, those in the sound condition donated 165.90 yuans in total, or 3.39 yuans per person,
whereas those in the no-sound condition donated 99.30
yuans in total, or 2.42 yuans per person. Because we did
not know the exact types of coins each donor donated, we
were unable to run significance tests on these results.
However, we knew the exact number of coins each donor
donated, and the distributions of the different types of coins
did not differ significantly between the two conditions.
Therefore, we used the number of coins to compare the
two conditions. As table 1 shows, those in the sound condition donated coins significantly more (M ¼ 3.92,
SD ¼ 2.69) than those in the no-sound condition (M ¼ 2.76,
SD ¼ 1.71; F(1, 88) ¼ 5.70, p < .05). (Even if we included the nondonors, the result was still significant,
F(1, 390) ¼ 6.33, p < .01.)
A possible alternative explanation for the finding was
that donors in the sound condition wanted to use the sound
to signal to others (the fundraiser and other customers in
the shop) that they were generous. Although we could not
completely rule out this explanation, we do not find it
11
viable. The fundraiser was always nearby and could see
how many coins one donated without the sound. The other
customers were seated far from the cashier, and the coffee
shop was full of ambient sounds such as coffee grinding
noise and background music so that it would be difficult
for the other customers to notice the sound of the donation
box. Additionally, this alternative explanation would not
apply to the other studies reported in this article and therefore is not a parsimonious explanation. Study 8 replicated
study 2 (pay what you want) in the field, and it showed that
the mere presence of a sound significantly bolstered behaviors that involved real monetary consequences.
To test the robustness of our findings, we also performed
a meta-analysis (Winer 1971) comparing the no-reaction
condition and the reaction (contingent-stimulus) condition
of all the studies, except for study 5 (which did not have a
behavioral measure) and study 7 (which adopted a withinparticipants design). The analysis yielded a highly significant effect (z ¼ 15.30, p < .001).
55
60
65
70
GENERAL DISCUSSION
People like reactions and like to repeat behaviors that
generate reactions. People like to throw stones at ponds
that generate splashes; people like to tell jokes to those
who laugh afterward; people like to work on exercise machines that offer performance feedback. In many situations,
reactions are positively valenced, convey useful information, or both, and therefore people seek them by engaging
in reaction-generating behaviors. Yet the preference for reactions seems so deeply rooted that people will seek reactions even when the reactions are a priori nonpositive and
noninformative. Reactions per se are reinforcing. In order
for an a priori nonpositive stimulus to be reinforcing, it
must be contingent on the corresponding behavior and perceived to be a reaction. At the same time, by being contingent on the behavior and serving as a reaction, an a priori
nonpositive stimulus can itself become positive.
Open Questions and Speculations
The present research has left many questions unanswered that we hope future research will address. One
such question concerns the relationship between the reaction effect and stimulation seeking. When bored, humans
and other animals alike will seek stimulation, even negative stimulation (e.g., Hebb 1958; Jones et al. 1961; Wilson
et al. 2014; see also Hsee et al. 2010 for a related phenomenon). For example, humans confined in darkness would repeatedly press a button to seek illuminations (Jones et al.
1961), and rats restrained in a cage would even seek electric shocks as stimulation (Coppock 1954). We speculate
that stimulation seeking can augment the reaction effect
but is not necessary for the effect. For example, in study 2
(pay what you want) and study 8 (donation) of the current
75
80
85
90
95
100
12
5
10
15
20
25
30
35
40
45
50
55
research, participants were not required to stay in the experiment and would not have felt restrained or bored;
therefore, boredom and stimulation seeking could not explain the reaction effect observed in those studies.
Another open question asks whether the reaction effect
is contradictory to operant conditioning. In a narrow
sense, the reaction effect (that even a priori negative reaction can be reinforcing) is at variance with operant conditioning. But in a broader sense, the reaction effect may be
in line with conditioning. An a priori negative stimulus
first becomes positive by serving as a reaction, namely,
by being associated with the completion of an action
(classical conditioning); then it proceeds to reinforce the
action that has made it positive in the first place (operant
conditioning). This speculation predicts a temporal sequence of first classical conditioning and then operant
conditioning. Indeed, prior studies have documented that
classically conditioned stimuli can serve as reinforcers in
operant conditioning (e.g., Ayllon and Azrin 1966; Holz
and Azrin 1962; Kelleher and Morse 1968). For instance,
Holz and Azrin (1961) found that electric shocks can reinforce the pecking behavior of pigeons (operant conditioning) if the shocks are first paired with food (classical
conditioning). However, if the reaction effect in our studies involved any classical conditioning, it was a subtle
form of classical conditioning. Unlike the electric shocks
in Holz and Azrin’s study, which were paired with food,
the negative reactions in our studies were not associated
with any external rewards but associated only with one’s
actions.
A third open question concerns the notion of contingency, which has different types. One type describes
whether the occurrence of a stimulus is contingent on the
occurrence of a behavior, for example, whether the emission of a sound immediately follows a jump. Another type
of contingency refers to whether the magnitude of a stimulus is contingent on the magnitude of a behavior, for example, whether the loudness of the sound depends on the
height of the jump. In this research, we have focused only
on the first type of contingency. We suspect that the second
type is also important in determining the reinforcing value
of a stimulus. All else being equal, a stimulus may be more
reinforcing if its magnitude is contingent on the magnitude
of the behavior than if it is not.
Since our research has focused on negative reactions,
readers may ask whether the arousing nature of negative
stimuli could explain the reaction effect (e.g., Baumeister
et al. 2001; Fiske 1980; Ohira, Winton, and Oyama 1998;
see also Travers, Reid, and Wagenen 1963 for a review). It
is possible that participants in the negative-reaction condition of our studies were more aroused and hence repeated
the activity more times than participants in the no-reaction
condition. But arousal could not explain parsimoniously
why they repeated the task fewer times than participants in
the positive reaction condition. Nor could arousal explain
JOURNAL OF CONSUMER RESEARCH
the difference between the contingent and the noncontingent negative stimuli conditions.
Readers may also wonder about the durability of the reaction effect. The studies reported in this research demonstrate the effect only in short tasks. Can the effect last
long? We suspect that for reactions to produce a longlasting effect, the reactions may have to be uncertain or
varied, rather than certain and fixed, so as to evoke curiosity, hinder hedonic habituation, or both.
60
Implications for Consumer Behavior
65
The current research joins a growing body of literature
(e.g., Thaler and Sunstein 2008; see also Gneezy et al.
2010; Hsee et al. 2013a,b; Jain et al. 2013; Larrick and Soll
2008; Soman and Cheema 2011; Yang et al. 2013; Zhao,
Lee, and Soman 2012) by demonstrating that subtle psychological manipulations can exert significant influences
on behaviors. In our case, the manipulation is a reaction.
Many practitioners have already realized the importance
of reactions and designed reaction-generating products. For
example, treadmill manufacturers have designed their
equipment so that its users can see reinforcing and useful
numbers (such as miles traveled) on the display panel as
they run on the machine. Software companies have developed applications that track the number of times one has
walked or done pushups. Vending machine manufacturers
have designed gumball machines with elaborate mechanisms that dispense the gumball through intricate ramps after one inserts a coin. Credit card companies have
introduced loyalty programs that award gift-redeemable
points every time the card holder makes a purchase. The
automobile maker Volkswagen has even designed stairs
and recycling bins that produce attractive sounds to induce
people to walk more and recycle more. In most of these examples, practitioners are taking advantage not only of the
effect of reactions, but also of the effect of positive valence
and the effect of useful feedback.
Our research suggests that practitioners can do even
more. For example, the numbers displayed on many treadmills do not feel like reactions because the numbers (e.g.,
miles traveled) change automatically as one runs and are
not contingent on one’s leg movement. To make the treadmill more motivating, designers may add a salient number
on the display panel that changes every time or every few
times one takes a stride (a feature that most treadmills
lack). Similarly, the numbers displayed on many personal
training apps do not feel like reactions either because the
numbers (e.g., a pushup count) change only silently and do
not catch the attention of the users as they are doing pushups. To make the app more engaging, developers may let
the software generate a sound every time (or every few
times) one does a pushup. As another example, the points
offered by many credit card companies also do not feel like
reactions because the points are added to one’s account
70
75
80
85
90
95
100
105
HSEE ET AL.
5
10
15
20
25
30
35
40
long after a purchase and the consumer may not even notice when the points are added. To increase usage, credit
card companies may send users a text message immediately after a purchase indicating that the points in their account have increased.
Whereas in the preceding examples we suggested practitioners can do more than what they are currently doing, in
other cases we suggest they can do less. For example,
vending machine manufacturers may not need to incur
extra costs to install elaborate mechanisms in gumball
machines, and exercise machine manufacturers may not
need to spend extra money to build fancy displays.
Machines that generate a simple and immediate reaction to
one’s action may be enough to reinforce the action.
Our research is also relevant to customer service.
Existing research finds that interactions with consumers
can increase their trust (Ramsey and Sohi 1997; Weitz,
Sujan, and Sujan 1986). Even measuring the satisfaction of
already satisfied consumers can bolster their purchase behavior (Dholakia and Morwitz 2002). Our research corroborates the finding by showing that even nonpositive
reactions can be rewarding and reinforcing. However, this
effect can be a double-edged sword when it comes to a pestering customer who keeps nagging a busy salesperson. We
predict that relative to no reaction, any reactions from the
salesperson (that are not highly aversive), even a grimace,
will make the customer feel good but at the same time will
encourage the consumer to keep nagging.
Marketers can also take advantage of the valence-change
effect identified in this research to improve consumer attitude toward their brands or logos. For example, currently
the logos of most stores are static. To increase consumer
preference, companies may design the logo on their web
pages so that online shoppers can “play with it,” for example, click on it to enlarge or shrink its size. Companies may
also design the logo in front of their physical stores so that
the neon light of the logo will flash every time a shopper
takes a step toward the store. Based on our valence-change
proposition, we expect consumers to like the logo more if
it reacts to their actions.
Coda
45
Humans are reaction-seeking animals. Among the myriad of reasons we do what we do, an important one may be
that our actions spur reactions. These reactions are splashes
in life, and these splashes keep us engaged and keep us
moving.
DATA COLLECTION INFORMATION
50
All the studies were designed jointly by all the authors.
Study 1 was conducted by research assistants under the supervision of the first author in summer 2014 in China.
Study 2 was conducted by research assistants under the
13
supervision of the first author in winter 2015 in the United
States. Study 3 was conducted by research assistants under
the supervision of the second author in summer 2013 in
China. Study 4 was conducted on MTurk by the third author in summer 2013 in the United States. Studies 5
through 7 were conducted on MTurk by the third author in
autumn 2014 in the United States. Study 8 was conducted
by research assistants under the supervision of all the authors in summer 2013 in China. All data were analyzed
jointly by the second and the third authors.
55
60
REFERENCES
Amir, On and Dan Ariely (2007), “Decisions by Rules: The Case
of Unwillingness to Pay for Beneficial Delays,” Journal of
Marketing Research, 44 (1), 142–52.
Ammons, R. B. (1956), “Effects of Knowledge of Performance: A
Survey and Tentative Theoretical Formulation,” Journal of
General Psychology, 54, 279–99.
Arkes, Hal R. and Peter Ayton (1999), “The Sunk Cost and
Concorde Effects: Are Humans Less Rational Than Lower
Animals?” Psychological Bulletin, 125 (5), 591–600.
Asch, David A., James P. Patton, and John C. Hershey (1990),
“Knowing for the Sake of Knowing: The Value of Prognostic
Information,” Medical Decision Making, 10 (1), 47–57.
Ayllon, T. and N. H. Azrin (1966), “Punishment as a
Discriminative Stimulus and Conditioned Reinforcer with
Humans,” Journal of the Experimental Analysis of Behavior,
9 (4), 411–19.
Bandura, Albert (1997), Self-Efficacy: The Exercise of Control,
New York: Worth Publishers.
Bangert-Drowns, Robert L., Chen-Lin C. Kulik, James A. Kulik,
and MaryTeresa Morgan (1991), “The Instructional Effect of
Feedback in Test-like Events,” Review of Educational
Research, 61 (2), 213–38.
Baron, Jonathan (2000), “The Effects of Overgeneralization on
Public Policy,” working paper, University of Pennsylvania.
Bastardi, Anthony and Eldar Shafir (1998), “On the Pursuit and
Misuse of Useless Information,” Journal of Personality and
Social Psychology, 75 (1), 19–32.
Baumeister, Roy F. (1998), “The Self,” in The Handbook of Social
Psychology, Vols. 1 and 2, 4th ed., ed. D. T. Gilbert, S. T.
Fiske, and G. Lindzey, New York: McGraw-Hill.
Baumeister, Roy F., Ellen Bratslavsky, Catrin Finkenauer, and
Kathleen D. Vohs (2001), “Bad Is Stronger Than Good,”
Review of General Psychology, 5 (4), 323–70.
Berlyne, Daniel E. (1966), “Curiosity and Exploration,” Science,
153 (3731), 25–33.
Boerke, Kyle W. and David Reitman (2011), “Token Economies,”
in Handbook of Applied Behavior Analysis, ed. Warne W.
Fisher, Cathleen C. Piazza, and Henry S. Roane, New York:
Guilford Press, 370–82.
Burger, Jerry M. (1992), Desire for Control: Personality, Social,
and Clinical Perspectives, New York: Plenum Press.
Burger, Jerry M. and Harris M. Cooper (1979), “The Desirability
of Control,” Motivation and Emotion, 3 (4), 381–93.
Coppock, Harold W. (1954), “Stimuli Preceding Electric
Shock Can Acquire Positive Reinforcing Properties,”
Journal of Comparative and Physiological Psychology, 47
(2), 109–13.
65
70
75
80
85
90
95
100
105
14
5
10
15
20
25
30
35
40
45
50
55
60
Deci, Edward L. and Richard M. Ryan (2000), “The What”
and “Why of Goal Pursuits: Human Needs and the
Self-Determination of Behavior,” Psychological Inquiry, 11
(4), 227–68.
Dervin, Brenda (1998), “Sense-making Theory and Practice: An
Overview of User Interests in Knowledge Seeking and Use,”
Journal of Knowledge Management, 2 (2), 36–46.
Dholakia, Utpal M. and Vicki G. Morwitz (2002), “The Scope and
Persistence of Mere Measurement Effects: Evidence from a
Field Study of Customer Satisfaction Measurement,” Journal
of Consumer Research, 29 (2), 159–67.
Fang, Xiang, Surendra Singh, and Rohini Ahluwalia (2007),
“An Examination of Different Explanations for the
Mere Exposure Effect,” Journal of Consumer Research, 34
(1), 97–103.
Ferster, Charles B. and Burrhus F. Skinner (1957), Schedules of
Reinforcement, East Norwalk, CT: Appleton-Century-Crofts.
Finkelstein, Stacey R. and Ayelet Fishbach (2012), “Tell Me What
I Did Wrong: Experts Seek and Respond to Negative
Feedback,” Journal of Consumer Research, 39 (1), 22–38.
Fiske, Susan T. (1980), “Attention and Weight in Person
Perception: The Impact of Negative and Extreme
Behavior,” Journal of Personality and Social Psychology,
38 (6), 889–906.
Girdner, John B. (1953), “An Experimental Analysis of the
Behavioral Effects of a Perceptual Consequence Unrelated
to Organic Drive States,” American Psychologist, 8 (8),
354–55.
Gneezy, Ayelet, Uri Gneezy, Leif D. Nelson, and Amber Brown
(2010), “Shared Social Responsibility: A Field Experiment in
Pay-What-You-Want Pricing and Charitable Giving,”
Science, 329 (5989), 325–27.
Golman, Russell, and George Loewenstein (2012), “Curiosity,
Information Gaps, and the Utility of Knowledge,” working
paper, Carnegie Mellon University.
Greenspoon, Joel (1954), “The Effect of Two Nonverbal Stimuli
on the Frequency of Members of Two Verbal Response
Classes,” American Psychologist, 9, 384.
Groos, Karl (1901), The Play of Man, New York: D. Appleton.
Hamerman, Eric J. and Gita V. Johar (2013), “Conditioned
Superstition: Desire for Control and Consumer Brand
Preferences,” Journal of Consumer Research, 40 (3), 428–43.
Hattie, John and Helen Timperley (2007), “The Power of
Feedback,” Review of Educational Research, 77 (1), 81–112.
Hebb, Donald O. (1958), “The Motivating Effects of
Exteroceptive Stimulation,” American Psychologist, 13 (3),
109–13.
Holz, William C. and Nathan H. Azrin (1961), “Discriminative
Properties of Punishment,” Journal of the Experimental
Analysis of Behavior, 4 (3), 225–32.
———. (1962), “Interactions between the Discriminative and
Aversive Properties of Punishment,” Journal of the
Experimental Analysis of Behavior, 5 (2), 229–34.
Hsee, Christopher K., Adelle X. Yang, and Liangyan Wang
(2010), “Idleness Aversion and the Need for Justifiable
Busyness,” Psychological Science, 21 (7), 926–30.
Hsee, Christopher K., Fang Yu, Jiao Zhang, and Yan Zhang
(2003), “Medium Maximization,” Journal of Consumer
Research, 30 (1), 1–14.
Hsee, Christopher K., Jiao Zhang, Liangyan Wang, and Shirley
Zhang (2013a), Magnitude, time and risk differ similarly
between joint and single evaluations Journal of Consumer
Research, 40 (June), 172–184.
JOURNAL OF CONSUMER RESEARCH
Hsee, Christopher K., Jiao Zhang, Zoe Y. Lu, and Fei Xu (2013b),
Unit-asking: A method to boost donations and beyond
Psychological Science, 24 (9), 1801–1808.
Hurwitz, H. M. B. (1956), “Conditioned Responses in Rats
Reinforced by Light,” British Journal of Animal Behaviour, 4
(1), 31–33.
Inesi, M. Ena, Simona Botti, David Dubois, Derek D. Rucker, and
Adam D. Galinsky (2011), “Power and Choice: Their
Dynamic Interplay in Quenching the Thirst for Personal
Control,” Psychological Science, 22 (8), 1042–48.
Jain, Kriti, Kanchan Mukherjee, Neil J. Bearden, and Anil Gaba
(2013), “Unpacking the Future: A Nudge Toward Wider
Subjective Confidence Intervals,” Management Science, 59
(9), 1970–87.
Jones, Austin, H., Jean Wilkinson, and Ina Braden (1961),
“Information Deprivation as a Motivational Variable,”
Journal of Experimental Psychology, 62 (2), 126–37.
Kazdin, Alan E. (1982), “The Token Economy: A Decade Later,”
Journal of Applied Behavior Analysis, 15 (3), 431–45.
Kelleher, R. T. and W. H. Morse (1968), “Schedules Using
Noxious Stimuli: III. Responding Maintained with Responseproduced Electric Shocks,” Journal of the Experimental
Analysis of Behavior, 11 (6), 819–38.
Kim, Ju-Young, Martin Natter, and Martin Spann (2009), “Pay
What You Want: A New Participative Pricing Mechanism,”
Journal of Marketing, 73 (1), 44–58.
Kling, J. W., L. Horowitz, and J. E. Delhagen (1956), “Light as a
Positive Reinforcer for Rat Responding,” Psychological
Reports, 2, 337–40.
Kuhnen, Camelia M. and Agnieszka Tymula (2012), “Feedback,
Self-esteem, and Performance in Organizations,” Management
Science, 58 (1), 94–113.
Larrick, Richard P. and Jack B. Soll (2008), “The MPG Illusion,”
Science, 320 (5883), 1593–94.
Levy, Paul E., Michelle D. Albright, Brian D. Cawley, and Jane R.
Williams (1995), “Situational and Individual Determinants of
Feedback Seeking: A Closer Look at the Process,”
Organizational Behavior and Human Decision Processes, 62
(1), 23–37.
Locke, Edwin A., Norman Cartledge, and Jeffrey Koeppel (1968),
“Motivational Effects of Knowledge of Results: A Goal-setting Phenomenon?” Psychological Bulletin, 70 (6, Pt. 1),
474–85.
Loewenstein, George (1994), “The Psychology of Curiosity: A
Review and Reinterpretation,” Psychological Bulletin, 116
(1), 75–98.
London, Manuel and James W. Smither (2002), “Feedback
Orientation, Feedback Culture, and the Longitudinal
Performance Management Process,” Human Resource
Management Review, 12 (1), 81–100.
Montgomery, Kay C. (1952), “Exploratory Behavior and Its
Relation to Spontaneous Alternation in a Series of Maze
Exposures, “ Journal of Comparative and Physiological
Psychology, 45 (1), 50–7.
Muenzinger, Karl F. (1934), “Motivation in Learning. I. Electric
Shock for Correct Response in the Visual Discrimination
Habit,” Journal of Comparative Psychology, 17 (2), 267–77.
Nunes, Joseph C. and Xavier Drèze (2006), “The Endowed
Progress Effect: How Artificial Advancement Increases
Effort,” Journal of Consumer Research, 32 (4), 504–12.
Nuthmann, Anne M. (1957), “Conditioning of a Response Class
on a Personality Test,” Journal of Abnormal and Social
Psychology, 54 (1), 19–23.
65
70
75
80
85
90
95
100
105
110
115
120
125
HSEE ET AL.
5
10
15
20
25
30
35
40
45
50
55
Ohira, Hideki, Ward M. Winton, and Makiko Oyama (1998),
“Effects of Stimulus Valence on Recognition Memory and
Endogenous Eyeblinks: Further Evidence for PositiveNegative Asymmetry,” Personality and Social Psychology
Bulletin, 24 (9), 986–93.
Piaget, Jean (1952), The Origins of Intelligence in Children, New
York: International Universities Press.
———. (1969), The Psychology of Intelligence, Totowa, NJ:
Littlefield Adams.
Preacher, Kristopher J. and Andrew F. Hayes (2008), “Asymptotic
and Resampling Strategies for Assessing and Comparing
Indirect Effects in Multiple Mediator Models,” Behavior
Research Methods, 40 (3), 879–91.
Ramsey, Rosemary P. and Ravipreet S. Sohi (1997), “Listening to
Your Customers: The Impact of Perceived Salesperson
Listening Behavior on Relationship Outcomes,” Journal of
the Academy of Marketing Science, 25 (2), 127–37.
Ratchford, Brian T., Debabrata Talukdar, and Myung-Soo Lee
(2007), “The Impact of the Internet on Consumers’ Use of
Information Sources for Automobiles: A Re-Inquiry,”
Journal of Consumer Research, 34 (1), 111–19.
Roberts, Carl L., Melvin H. Marx, and George Collier (1958),
“Light Onset and Light Offset as Reinforcers for the Albino
Rat,” Journal of Comparative and Physiological Psychology,
51 (5), 575–79.
Rodin, Judith and Ellen J. Langer (1976), “The Effect of Choice
and Enhanced Personal Responsibility for the Aged: A Field
Experiment in an Institutional Setting,” Journal of
Personality and Social Psychology, 34 (2), 191–98.
Rucker, Derek D. and Adam D. Galinsky (2008), “Desire to
Acquire: Powerlessness and Compensatory Consumption,”
Journal of Consumer Research, 35 (2), 257–67.
Ryan, Richard M., C. Scott Rigby, and Andrew Przybylski (2006),
“The Motivational Pull of Video Games: A Selfdetermination Theory Approach,” Motivation and Emotion,
30 (4), 344–60.
Shute, Valerie J. (2008), “Focus on Formative Feedback,” Review
of Educational Research, 78 (1), 153–89.
Sidowski, Joseph B. (1954), “Influence of Awareness of
Reinforcement on Verbal Conditioning,” Journal of
Experimental Psychology, 48 (5), 355–60.
Skinner, Burrhus F. (1969), Contingencies of Reinforcement: A
Theoretical Analysis, New York: Appleton-Century-Crofts.
———. (1953), Science and Human Behavior, New York:
Macmillan.
———. (1957), Verbal Behavior, New York: Appleton-CenturyCrofts.
Smither, James W., Manuel London, and Richard R. Reilly (2005),
“Does Performance Improve Following Multisource
Feedback? A Theoretical Model, Meta-Analysis, and
Review of Empirical Findings,” Personnel Psychology, 58 (1),
33–66.
Soman, Dilip and Amar Cheema (2011), “Earmarking and
Partitioning: Increasing Saving by Low-income Households,”
Journal of Marketing Research, 48 (Special Issue), S14-S22.
15
Sweetser, Penelope and Peta Wyeth (2005), “GameFlow: A
Model for Evaluating Player Enjoyment in Games,” ACM
Computers in Entertainment, 3 (3), 1–24.
Taffel, Charles (1955), “Anxiety and the Conditioning of Verbal
Behavior,” Journal of Abnormal and Social Psychology, 51
(3), 496–501.
Thaler, Richard H. and Cass R. Sunstein (2008), Nudge:
Improving Decisions about Health, Wealth, and Happiness,
New Haven: Yale University Press.
Thorndike, Edward L. (1898), “Animal Intelligence: An
Experimental Study of the Associative Processes in
Animals,” Psychological Monographs: General and Applied,
2 (4), i–109.
Travers, Robert M. W., Ian E. Reid, and R. Keith Van Wagenen
(1963), Reinforcement: A Review of Selected Research, Salt
Lake City: University of Utah, Bureau of Educational
Research.
Tolman, Edward C., Calvin S. Hall, and E. P. Bretnall (1932), “A
Disproof of the Law of Effect and a Substitution of the Laws
of Emphasis, Motivation and Disruption,” Journal of
Experimental Psychology, 15 (6), 601–614.
Verplanck, William S. (1956), “The Operant Conditioning
of Human Motor Behavior,” Psychological Bulletin, 53 (1),
70–83.
Weitz, Barton A., Harish Sujan, and Mita Sujan (1986),
“Knowledge, Motivation, and Adaptive Behavior: A
Framework for Improving Selling Effectiveness,” Journal of
Marketing, 50, (October), 174-91.
White, Robert W. (1959), “Motivation Reconsidered: The
Concept of Competence,” Psychological Review, 66 (5),
297–333.
Whitson, Jennifer A. and Adam D. Galinsky (2008),
“Lacking Control Increases Illusory Pattern Perception,”
Science, 322 (5898), 115–17.
Wilson, Timothy D., David A. Reinhard, Erin C. Westgate, Daniel
T. Gilbert, Nicole Ellerbeck, Cheryl Hahn, Casey L. Brown,
and Adi Shaked (2014), “Just Think: The Challenges of the
Disengaged Mind,” Science, 345 (6192), 75–77.
Winer, Ben James (1971), Statistical Principles in Experimental
Design, 2nd ed., New York: McGraw-Hill.
Yang, Yang, Joachim Vosgerau, and George Loewenstein (2013),
“Framing Influences Willingness to Pay but Not
Willingness to Accept,” Journal of Marketing Research, 50 (
6), 725–738.
Zajonc, Robert B. (1968), “Attitudinal Effects of Mere Exposure,”
Journal of Personality and Social Psychology, 9 (2p2), 1–27.
Zhang, Ying and Szu-Chi Huang (2010), “How Endowed
versus Earned Progress Impacts Consumer Goal
Commitment and Motivation,” Journal of Consumer
Research, 37 (4), 641–54.
Zhao, Min, Leonard Lee, and Dilip Soman (2012), “Crossing the
Virtual Boundary The Effect of Task-Irrelevant
Environmental Cues on Task Implementation,” Psychological
Science, 23 (10), 1200–07.
60
65
70
75
80
85
90
95
100
105